Learning Complex Spatial Relation Model from Spatial Data

Sheng-Sheng Wang ； Ru-Yi Dong

《電腦學刊》 30卷6期 (2019/12) Pp. 123-136

https://doi.org/10.3966/199115992019123006010

摘要

Describing spatial relations is a challenging task in image understanding and content-based image retrieval. There are only few works focused on describing complex spatial relations, and they usually use mathematical system which is not self-adapted, i.e. one model is only suitable for certain group of datasets. To address these problems, we proposed a self-adapted method for describing complex spatial relations. With our method, the complex spatial relation model can be quickly and accurately generated from very few labeled samples without priori-knowledge. The proposed method is tested on several benchmark datasets, and the experiment results demonstrate the superior performance and the robustness of our method.

關鍵字

classification ； complex spatial relations ； feature selection ； qualitative spatial reasoning